k8s-network-troubleshooting
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources like Kubernetes logs and object descriptions which could theoretically contain malicious instructions if a workload is compromised.
- Ingestion points:
kubectl logs,kubectl describe(found inSKILL.md,references/dns.md, andreferences/services.md). - Boundary markers: The skill does not define specific delimiters for log output, but instructions focus the agent on identifying network-specific error patterns.
- Capability inventory: The skill has access to
kubectl execandkubectl patch, which are powerful cluster management operations. - Sanitization: The skill implements human-in-the-loop control by requiring the agent to propose fixes and ask for permission before running state-changing commands.
- [COMMAND_EXECUTION]: The skill uses
kubectl execto perform connectivity tests inside user pods, which allows for command execution within the container environment. - Evidence:
kubectl exec deployment/my-web -n my-other-namespace -- nslookup my-app.my-namespaceinSKILL.md. - Mitigation: All such commands are gated behind an 'Ask first' directive, ensuring the agent cannot execute them without explicit user consent.
- [SAFE]: The skill proactively implements security best practices by instructing the agent to never read Kubernetes Secrets and to redact sensitive information like tokens or passwords from its output.
Audit Metadata